Tenstorrent Unveils Next-Gen Servers for Fast Tokens, No Disaggregation Needed

Find the Latest SoC Solutions
for... Automotive... IoT... Security... Audio... Video

Apr. 30, 2026 –

“The world is talking about specialized, specialized, specialized, and that should be terrifying them because as models change, that specialized hardware is not going to work,” said Tenstorrent CEO Jim Keller.

By Sally Ward-Foxton, EE Times

Against a growing industry trend for disaggregated inference, Tenstorrent will unveil its Galaxy Blackhole server and cluster offering later this week, which is designed to offer fast token generation and efficient tokenomics from only Tenstorrent chips.

The new 6U servers offer 23 PFLOPS (Block FP8) from 32 Tenstorrent Blackhole chips, which handle both prefill and decode. Typical customers are deploying 4- to 32-Galaxy clusters today, but bigger systems are also in the works, Tenstorrent CEO Jim Keller told EE Times.

“We started making the Galaxy Blackhole servers in January, and somewhere along the way, we started to realize just how fast Tenstorrent AI is,” Keller said. “We do something no one else does, where we can hook up a lot of, let’s say, medium-performance chips together in a Galaxy box, then hook those together, and scale applications across multiple Galaxies.”

Click here to read more ...

 Back